During motion estimation between two frames of a sequence of frames there sometimes is the problem that certain areas of one frame are not visible in the other frame. Typically this is the case for pixels belonging to an object in the background, which in one of the frames are hidden by pixels belonging to a foreground object. In such cases it is not possible to determine a motion vector for the occluded pixels directly from the two frames. Most of today's motion estimators simply assume that the pixels are present in the other frame and provide motion vectors for these pixels, which optimize a selected criterion. However, these motion vectors do generally not correspond to the correct motion vectors.